I know you’ve all been waiting for it. We’ve talked about putting ants on stilts, kidnapping baby gerbils, and hijacking a truck full of geese. All in the name of science. Ants and gerbils taught us about the limitations of the path integration system, but also how amazingly cool it is. The geese suggested that path integration is probably innate. We took a brief digression to talk about how to tell if some cognitive ability is innate, and the historical background to the nativism versus empiricism debate.

And now, humans. Buckle your seatbelts. This is going to be a wild ride.

It’s clear that adult humans and even older children have extensive spatial knowledge. Consider that the ability of children to find shortcuts through scary forests requires extensive understanding of the spatial layout of their environments, and of the body’s position within the environment. If they’re lucky, they don’t stumble upon a house made of candy.

Figure 1: I’d live there. I’d kick out the mean old lady though. She’s no fun. My room can be the one made of bacon and chocolate. Sci isn’t allowed there. She doesn’t like bacon with chocolate.

Figure 2: Yum.

But I digress.

The fact that human adults (and animals!) have these abilities raise two important questions:
(1) What is the nature of the early mental representations of space, from which more sophisticated representations develop?
(2) What are the environmental conditions under which this system of knowledge develops?

Read about the experiments behind the cut!

Meet Kelli. Kelli is 2 years old, and is the first-born of two children. Her sister is 12 months younger than she is. She is the surviving member of a pair of twins who were born approximately three months prematurely. Since Kelli was so sick as a newborn, weighing in at only slightly more than two pounds, she remained in the hospital for the first six months of life. When she was sent home from the hospital with a weight of six pounds, she had the developmental status as a newborn (if she had been born on time – 3 months later – she’d be considered a 3-month-old newborn).

During the first three weeks of her life, Kelli became a victim of Retrolental Fibroplasia (it is now called Retinopathy of Prematurity). RLF is judged in terms of grades 1 through 5. Grades 1-3 leave some vision and can be reversible. Grade 4 leaves no vision but possibly some basic light perception. Grade 5 leaves no vision or light perception. Kelli had grade 5 RLF in both eyes, and is considered totally blind by her opthalmologist. When she became 4 years old, she appeared to notice some light from a lamp in her right eye at a distance of less than two feet, but her doctor still considered her totally blind.

By 21 months of age, Kelli could locate objects by sound, and could recognize them with her mouth, fingers, or by scent. She could correctly identify more than a dozen of her own body parts, and could find the matching body parts on her mother. She knew the physical configuration of everyday objects. For example, she could manipulate a hairbrush by finding the handle, and using the bristled end in her hair. She execute different motor strategies (with her hands and arms) to manipulate and retrieve objects from containers, depending on the structure of the container, such as a plate or a bowl. She executed different motor sequences to climb on different kinds of furniture in her house, and seemed to be able to anticipate subsequent movements. For example, she climbed backwards up her highchair to sit in the seat, rather than flipping over to sit on the couch. She knew which cabinets in the kitchen she was permitted to open, and which she was not. By thirty months, she could navigate successfully between different rooms in her house.

This is all to say that Kelli had a fairly sophisticated knowledge of the physical and spatial structure of objects and her world, and the relation of her body to the outside world. What was unclear was whether or not this knowledge was based on learned sequences of actions, or not.

By now, you probably know where this is going. Children best discover the spatial organization of the world by looking and acting upon the world. Blind kids obviously can’t use vision to help, and they also reach, crawl, and walk (which is to say, act upon the world) much later in development than sighted kids. You might, therefore, expect that blind kids have significantly less experience with the spatial properties of the world. If spatial knowledge is demonstrated by a young blind child, this system of knowledge can develop under a wide range of environments, even under severe deprivation.

Let the experiments begin!

Experiment 1: Kelli is 31 months old.

Kelli is brought into a playroom, where she has never been before. Four separate objects were in the room: a chair, some pillows, a table, and a basket full of toys. Sounds like fun! After mom sat down in the chair, Kelli was shown around the room by the experimenter.

Figure 3: Layout of the room. Symbols indicate the location of (M)om, (P)illows, (T)able, and (B)asket of toys.

She was led along the routes indicated by the dotted lines. From Mom to pillows, and back to Mom, twice. From Mom to the table, and back to Mom, twice. Finally, from Mom to the basket, and back to Mom, twice. She was allowed to touch the objects each time.

The solid lines represent test routes. She had never traveled the route from table to basket, or from table to pillows, or from pillows to basket. If she successfully navigated the unfamiliar routes, then Kelli made spatial inferences based on a mental map of the room. If she simply retraced her steps along learned routes, then this might suggest that she did not have a mental representation of the spatial layout of the room.

So Kelli was led from Mom to the table. Then she was given directions, like “go to the toy basket” or “find the pillows.” The experimenter encouraged her throughout each trial, but did not interfere with Kelli’s movements. If Kelli was within 1 foot of the target object, the trial was considered a success, and she was praised for her success. If she became visibly confused, or stated explicitly that she couldn’t find the target, or if she approached an incorrect object, the trial was considered a failure.

Results? One point for Kelli. She did indeed make spatial inferences, finding new routes between the various objects. While they were not always straight lines, her initial angular estimate (the direction she turned her body at the beginning) was almost always such that if she could walk a perfectly straight line, it would have led her directly to the object.

Experiment 2: Kelli is 43 months old.

Kelli’s success in experiment 1 could have been because the objects served as landmarks.  Did she use the landmarks themselves to align herself with the goal object, or did the landmarks simply represent places in space? Experiment 2 explored this question.

The setup was the same, except two of the objects were removed, and thus two of the four locations were not occupied by an object. The arrangement of the spaces was similar, but the room was a different size. As in experiment 1, she was trained (twice in a row per route) on three different routes, and tested on the un-learned routes between them.

Results? Two points for Kelli. In 12 out of 12 trials, her initial angular estimate was always towards the location. Her final position was near the proper location on the majority of those trials.

Comparing the object and location trials, it seems that starting a route from a landmark may enhance performance, but only slightly. She was in the proper location on 4/5 object trials, and 6/8 location trials. Thus, object landmarks facilitate spatial orientation, but are not necessary for navigation.

Experiment 3: Kelli is 37 months old*

*The experiments are presented in logical order to explain the progression of experiments and ruling-out of alternatives, instead of in chronological order.

What might account for the errors seen previously? She erred in approximately one quarter of the trials. Possibilities:
(1) Deficiency in spatial knowledge, or in Kelli’s ability to properly utilize that knowledge.
(2) Problem in performance, rather than competence. Perhaps her errors were based on lack of attention, memory problems, distraction, and limited motor control.

To address this, Kelli’s task was to go to a place where she had heard a voice only a few seconds before. The only task, therefore, was to determine the location of the source of a sound and maintain it in memory while moving to the target. It was presented as a variation of the game “hide and seek,” and it was clear that she enjoyed the game (and therefore, was motivated to succeed). Her performance was similar to experiments 1 and 2, and it appeared that her difficulties concerned motor control or inexact locomotion, and not lack of applying spatial knowledge. She never took a straight path in this easier task. Perhaps she is easily distractable, like any young child. Perhaps it is difficult to maintain a straight line as a blind child. In any case, another point for Kelli.

Experiments 4, 5, and 6

These experiments were conducted to control for possibilities that Kelli used cues from inadvertent sound sources in the room, accidental cues from the experiment, or echolocation.

Experiment 4 (age 34 months) was identical to experiment 1, except in this case, after training, the spatial layout of the room was rotated 90 degrees. If there were external cues from the room that Kelli was using, they would lead her astray. This possibility was ruled out – she did not find her way by using external sound- or scent-related cues in the room. One more point for Kelli.

Experiment 5 (age 53 months) was identical to experiment 1, except that different experimenters were used in the training and testing phases. In training, Kelli was shown two identical wooden planks that served as “playing boards.” She was told that one was for her toys, and one was for her sister’s toys. The testing experimenter did not know which board Kelli had been told was her board, and which was her sister’s board. Therefore, when instructing Kelli to carry various toys to the various boards, the experimenter did not know which trials were successes and which were failures. Kelli succeeded on 8 out of 8 trials, confirming that she did not rely on subtle accidental cues from the experimenter.

Experiment 6 (age 48 months) was designed to investigate the possibility that Kelli used echolocation (like dolphins and bats!) to navigate the room. Some blind adults have been known to use a form of echolocation to navigate the world. This is unlikely for Kelli, especially given the results of experiment 4 – upon rotating the objects within the room, the pattern of echoes within the room would change. Still, experiment 6 was designed to explicitly address this issue. Kelli was carried into the playroom, where a single object was previously placed. She was placed onto the floor, standing, and was simply asked to find the object. Across 12 trials, four different objects were used, three times each. As you might expect, Kelli was very confused. Her initial angular estimate was accurate on only 4 out of 12 trials. She ended up near the object on only 2 out of 12 trials. Even when she did succeed (and then, only rarely), it was clear that this was by chance and random. One one trial, she actually complained “I can’t find it,” when she was standing directly next to the object. In a sense, another point for Kelli. She was better able to navigate using a mental representation of the room than by detecting echoes.

Taken together, experiments 4-6 confirm that Kelli’s performance utilized spatial knowledge, not sensory information.

Experiment 7: Sighted controls.

Experiment 7 was identical to experiment 1, but was conducted with sighted-but-blindfolded children (average age 3 years) and adults (college undergrads). The kids’ performance was on par with Kelli’s performance. On average, the kids successfully found the object 7.4 times out of 12. As expected, the adults outperformed the kids, finding the object an average of 10.8 times out of 12.

The ability to navigate without vision obviously improves with age, but by the third year the ability is already well-developed, even among the blind.

Still with me? We’ve only got one more experiment to go, and it is really really cool. Hang in there.

Experiment 8: Using explicit knowledge.

Kelli is now 4 and a half years old. She is brought into the playroom, which, by this time, is already familiar to her, and was seated in a chair. She was given a map of the room. She asked what a map was, and was told “it’s something that shows you how to find things in the room.” She was handed a piece of cardboard, on which two small wooden blocks were glued. One block represented her position within the room, and the second represented a toy basket placed elsewhere in the room. The “map” was placed on her lap in the proper orientation, so that the layout matched the layout of the room. After ensuring that Kelli understood that the map was a “pretend” version of the room, the trials began. New “maps” were given to Kelli when the location of the toy basket was changed. Another giant win for Kelli: she was accurate on 10 out of 11 trials.

She understood the concept of a map the very first time it was explained to her, and could use that concept to navigate around the room. So she represented spatial information, and used it to make inferences about the world. Super cool!

Okay, let’s wrap it up. What can we learn from this?

First, Kelli’s performance suggests three properties of spatial knowledge:
(1) Generativity – Just as a language learner can speak and understand sentences he or she has never heard or said before, a young navigator can find new routes never before traveled.
(2) Abstractness – Given that a young blind child can successfully navigate, the ability is likely to be independent of modality. The information used is necessarily spatial, and not visual, auditory, or kinesthetic.
(3) Metric geometry – Mental spatial knowledge is probably organized in a metric fashion, most likely a Euclidean representation of space. (Hey! There’s our old friend Descartes, again!) If an individual knows two of the angles and distances between three objects, he or she can calculate the third angle and distance, using principles of basic Euclidean coordinate geometry.

As has been noted, the question of spatial knowledge is not new. What is new is the evidence that spatial knowledge arises naturally in humans, with little training and with no visual experience. This basic knowledge guides our earliest attempts in life to explore the world.

Reference:
Landau, B., Spelke, E., & Gleitman, H. (1984). Spatial knowledge in a young blind child Cognition, 16 (3), 225-260. DOI: 10.1016/0010-0277(84)90029-5